In subbnad image coding, we have obtained the optimum frequency band partition method which can determine both the optimum partition pattern in 2-dimensional frequency domain and the op-timum combination of quantizers for subbands based on the rate-distortion sense. This paper presents the optimum space-frequency partition method for subband image coding in order to improve the coding per-formance of our previous method. The proposed method optimizes the pruning of coefficient pixels in each subband from the rate-distortion sense to remove the redundancy in 2-dimensional spatial domain adding to the optimum partition pattern and the optimum combination of quantizers in our previous method. Also, the optimization algorithm for three parameters at the lower computational costs is developed using the tree structure between band blocks which is a small region of 2-dimensional frequency domain. Our experimental results show that the PSNR of the proposed method is about 2[dB] higher than our previous method at the same bit rate. Moreover, the coding performances of the proposed method are superior to those of the wavelet packet in spite of the computational costs of the proposed method is lower than the wavelet packet.

We introduce a method to automatically synthesize a nightscape image from a daytime city aerial view. Based on the scotopic vision, we adjust the image luminance, darken the entire image and add to this a drawing processing in line with several estimates. We capture the characteristics of the daytime city aerial views and then use morphological image processing plus connected component labeling to locate the specific areas of walls and windows. In addition, we regard the lower luminance parts in daytime city image as the area of road and extract its contour information. For the drawing operation, we use the window area color composition ratios based on statistical analysis and then reproduce the window lighting though the shape of the estimated window area. Furthermore, we simulate a phenomenon known as diffuse reflection of the walls on the road. We demonstrate some experimental results of our plausible nightscape image generation.